CIP MOTION MANUAL TUNING PROCESS

Manual Tuning Process

Overview

This document is intended to give a short overview of a process flow to manually tune a servo axis
connected to a mechanism. This includes Manual Tune of the Load Observer, when configured as Load
Observer with Velocity Estimate.

The users of this document need to be familiar with Motion Control concepts and have knowledge of
Kinetix 6500/5500 families of Drives.

The users of this document need also to be familiar with Rockwell Studio 5000 environment,
ControlLogix, and CompactLogix controllers

For detailed information on Kinetix Drives Parameters and CIP Motion technology, please refer to the
following Rockwell Automation Publication:

- MOTION-AT005A-EN-P (Motion System Tuning)

- MOTION-RM003C-EN-P (Integrated Motion on the Ethernet/IP Network)

The manual tuning process described here follows the known “Inside-Out” tuning method and extends
to the Load Observer PI regulator.

Additionally, it considers the Inertia estimation trough the Autotune function of the drive as a necessary
verification of the Mechanical system response. In effect, performance of the Servo system are critically
affected by the mechanical load and the decision to perform a Manual tuning process. This can be time-
consuming and requires some expertise, and is based on combined evaluation of Performance
Requirements and Mechanical system response.
Initially determine the System Inertia using the Autotune function, then moving inside-out through
Observer Tuning, Velocity Loop Tuning and Position Loop tuning, finally returning the FeedForward
gains. During each step we isolate (or try to) the loop we are testing from the following one, to be able
to determine easily and steadily each loop parameter.

AUTOTUNE

OBSERVER ACCELERATION
TUNING FEEDFORWARD

VELOCITY LOOP VELOCITY

TUNING FEEDFORWARD

POSITION LOOP
TUNING
Understand Requirements: There is no
need to perform a full tuning process to
maximize system performance if the
application does not require it.
START TUNING
PROCESS
The performance of a servo system is
acceptable when it falls within
Specifications

Autotune: When dealing with a mechanical

system we can use the Autotune with “measure
inertia” enabled to test the mechanical system.
AUTOTUNE Giving a torque input nearly open loop to the
mechanics we can analyze the velocity feedback
response to understand what type of system is in
front of us

Rigid Load: When the Autotune Torque is

applied, the Load simply accelerates to reach
the target velocity. If the result is a triangular
Velocity feedback waveform then inertia
estimation from the process is considered
reliable. Expect a relatively easy tuning.

Slightly Compliant Load: Here the

acceleration segment is linear yet
deceleration is imperfect, this indicates a
slight uncoupling of the load. This can be
evaluated as slightly compliant or small
“decoupled inertia ratio”. Inertia estimation
process is considered reliable.
Disturbed Load: The acceleration segment is
no longer linear. Therefore either the inertia
of the system is varying throughout the move
or there is an external non-constant
force/torque in action.

Vertical Load + Varying Inertia: The load

must be held with Torque Offset before
applying Autotune process. Velocity is far
from being a straight line. High inertia
variations/Counter Torque variations during
the cycle. Velocity Feedback shows sign of
high frequency noise.

Compliant Load: When subject to a torque

input the load consistently oscillates at its
natural frequency. Inertia estimation can be
misunderstood from the Autotune process.
We can expect a challenging tuning process.
 UNDERSTAND
MECHANICAL SYSTEM

Autotune result GREEN = Autotune with

After evaluating the Autotune feedback Rigid load and Medium/High system
results together with Application response generally Ok
Requirement, we need to decide if going
through a manual tuning process is valid. Autotune result YELLOW = Autotune with
For most applications, a better result is
Rigid load and Medium system response
achieved by enabling the Load Observer
with Velocity Estimate, than selecting generally Ok
Loop Response and Load Coupling from
the Autotune tab in the Drive properties Autotune result RED = Possible Manual
and simply apply. Tuning Required

Do we need to go
through a manual tuning
process?

We cannot access the Observer Loop directly to tune its

response so we need to tune it indirectly, based on its
effect on the command signal and feedback response of
the Motion system. We need to deactivate the external
STEP 1: OBSERVER
Velocity and Position Loops and let only the Observer
MANUAL TUNE drive the load.

IR = From Autotune

Position Loop Bandwidth = 0

Position Integrator Bandwidth = 0
Set Load Observer Velocity FeedForward = 0%
with Velocity Estimate
Velocity Loop Bandwidth = 0
Option
Velocity Integrator Bandwidth = 0
Acceleration FeedForward = 100%

Low Pass Filter = 0

Notch Filter = 0
In cases of Low frequency resonance, do not use
Notch or Low Pass filter. In all other cases you can
Set Loop gains to start with the Notch Filter set to 1000 Hz

deactivate Loops
 Acceleration
Command via
With all loops effectively disabled, the
Load Observer is the only regulator FeedForward
acting on the system. It will integrate the Generate a velocity profile
Acceleration Command transmitted from (Use a MAM with
the Acceleration FeedForward into its
original velocity profile and try to adjust Trapezoidal Profile). Do not
the Acceleration Reference to follow the saturate Current Feedback
commanded profile.

Trend CommandVelocity vs ActualVelocity

It is useful to trend Actual Velocity Feedback when
Tune the Observer applying the Velocity Estimate option. In this case, the
gains to best follow Velocity Feedback signal is NOT the real signal but it is
an estimate. Despite this fact it is still the preferred
the Velocity Profile parameter to trend for this part of the manual tuning
process.

Tune the proportional

gain first, then the
integral gain

1 - Start with Load Observer Gains zeroed (or very little)

2 - Trend Actual Velocity (Not Velocity Feedback because Velocity Feedback will be manipulated by the Observer)
3 - Gently increase Observer Proportional Gain until velocity profile reaches the desired shape.
4 - Increase profile aggressiveness to worsen Observer task
If the Observer Action shows signs of instability, stop increasing gains and reduce Observer Bandwidth by 2/3
If the Velocity Feedback shows signs of high frequency noise, you can try to tune it out by gradually reducing the Low Pass Filter. This will not help in
cases of Low Frequency Resonance (Mechanical Compliance)
5 - Add Integral gains to improve stiffness. (Be aware of Mechanical Compliance). Avoid overshooting.

Do not saturate Drive Current at any point!

Tune out Hi-Frequency
Noise using Notch filter
and Low Pass Filter

Start with Low Gains in

Observer Proportional
Regulator

Increase Gains to match

Velocity Command or
Stop at first signs of
instability and reduce
Proportional Gains

Stiffen using a little

Observer integral gain
 STEP 2: Velocity Loop
Manual Tune

IR = From Autotune

Position Loop Bandwidth = 0

Position Integrator Bandwidth = 0
Velocity FeedForward = 100%
Set Loop gains to
deactivate Position Loop Velocity Integrator Bandwidth = 0
Acceleration FeedForward = 0

Keep Observer value as tuned prev. step

Keep Filters value as tuned prev. step

Generate a step
velocity profile using a
Time Cam

Velocity Step Command

Tune the Velocity via Feed Forward
Loop to best follow
the Velocity Profile

1 - Start with Velocity Proportional Gain zeroed (or very little)

2 - Trend Actual Velocity (Not Velocity Feedback because Velocity Feedback will be manipulated by the Observer) vs CommandVelocity
3 – Gently increase Velocity Proportional Gain until velocity profile reaches the desired shape with no excessive overshooting. (A little overshooting
when tuning velocity Loop may be acceptable)
4 - Do not saturate Drive Current
If the Axis shows signs of instability, stop increasing gains and reduce Velocity Bandwidth by 2/3

Creating a Velocity Step Command generates a Torque pulse that stresses Mechanical system. The harmonic content of a pulse includes all
frequencies. This allows the drive to highlight every resonance in the mechanical system. If the velocity command is not a step command this is no
longer true and the tuning process result will be less accurate. Having the Load Observer in place mitigates the step command effect because the
Observer tries to compensate for mechanical responses deviating from ideal behavior.
 Start with low Velocity
Bandwidth Gain

Gently increase Velocity

Loop Bandwidth to match
the Command Velocity

A little overshoot in tuning

the velocity loop may be
acceptable
 STEP 3: Position Loop
Manual Tune

IR = From Autotune

Velocity FeedForward = 0
Disable the Axis. Zero the Acceleration FeedForward = 0
FeedForwards and enable
Position Loop Gains Keep Observer value as tuned prev. step
Keep Filters value as tuned prev. step
Keep Velocity Loop Gains as tuned prev. step

Use Application
Motion Profile

Tune the Position

Loop to minimize
Position Error

1 - Start with Position Proportional Gain zeroed (or very little)

2 - Trend Actual Velocity (Not Velocity Feedback because Velocity Feedback will be manipulated by the Observer) vs
Command Velocity
3 – Trend Position Error
3 - Gently increase Position Proportional Gain to minimize Position error, still ensuring no overshoots in Velocity when
following the velocity command
Typically, if the Inertia Ratio is correctly measured, the Position Loop Gain does not exceed 2/3 of Velocity Loop
Bandwidth
4 - Do not saturate Drive Current
 Increase Position
Proportional Gain to
reduce position error.
Trend Velocity Command
and Actual Velocity

Do not accept
overshooting. It will
largely increases when re-
introducing Velocity
FeedForward

STEP 4: Back-in with By directly providing the Command

the Velocity and Value to the following Loop, the
response is much faster than waiting for
Acceleration the normal command signals, which in
FeedForwards turn results in a greatly reduced Position
and Velocity Error during movement.
 Gently increase the
Velocity FeedForward
until velocity response
shows overshooting or
you reach 100%

Try to reduce overshoot

by applying some
Acceleration
FeedForward. If this
happens you can iterate
this process increasing
both FeedForwards

Stop increasing the

Acceleration
FeedForward when it no
longer limits the
overshooting
 The FeedForwards greatly reduce
position error during movement, but
they do not add precision or accuracy
as that depends upon Loop Gains.